Accurate detection of Mineral Oil Saturated Hydrocarbons (MOSH) and Mineral Oil Aromatic Hydrocarbons (MOAH) is critical for food safety and regulatory compliance. However, inconsistencies in chromatographic integration, baseline corrections, and subjective analyst interpretation create variability, hindering standardization and reproducibility (JRC report: Robouch et al., 2022, JRC129603). This session introduces a transformative AI-driven solution: Generative Limited Sample Model AI, a novel approach designed to eliminate analytical variability and standardize MOSH/MOAH analysis across laboratories. Unlike conventional deep learning models that require extensive datasets, LSM AI learns from a limited set of expert-annotated samples, providing a scalable, high-precision, and adaptable solution for contaminant detection. This innovation allows laboratories to move beyond subjective, manual data processing and embrace fully automated chromatographic deconvolution, matrix effect correction, and AI-driven regulatory compliance workflows. Key Innovations and Analytical Advancements

AI-Driven Chromatographic Peak Deconvolution – Eliminates manual bias, ensuring reproducible MOSH/MOAH quantification across GC-FID and GC×GC-TOFMS platforms. Instrument Variation Compensation – Dynamically corrects for inter-laboratory discrepancies, harmonizing data processing with real-time LSM AI-driven calibration. MOSH/MOAH Source Attribution – Enhances regulatory compliance by classifying contamination origins (food processing lubricants, recycled packaging, environmental deposition). Automated Workflows for Non-Specialists – Expands adoption by replacing expert-dependent manual workflows with AI-powered data interpretation and compliance tracking. Regulatory Alignment & AOAC Standardization – Delivers standardized, reproducible, and bias-free quantification, supporting AOAC and JRC guidelines for method validation.

This breakthrough enables scalable, automated MOSH/MOAH analysis, ensuring regulatory-ready, high-accuracy results without manual inconsistencies.